An address comprises a state name, a city name and a street name. Thus, some navigation devices receive voice input of address elements in the order of: a state name, a city name and a street name, in order to specify a site (See Japanese Un-examined Patent Application, Laid-Open No. 2003-140682, which is hereinafter referred to as Patent Document 1). On the other hand, considering cases where a vast number of streets exist, some navigation devices request a user to utter two letters of a street name first. In detail, a user is invited to utter up to the second letter of a the street name, and those two uttered letters are recognized. Candidates for two recognized letters are displayed. Among the candidates, the user is invited to select two letters by voice, and street names beginning with the selected two letters are made candidates. Then, the user utters the whole name (the full name) of the street, and, among the candidates for the street name, one that best matches the utterance is extracted as the recognition result.